Jun Chen

College of Mechanical and Electronic Engineering, Northwest A&F University,
Yangling, Shaanxi, China

Fruit harvesting represents a highly seasonal and labor-intensive agricultural endeavor. The increasing labor costs, combined with relatively low market prices and a shortage of qualified and skilled workers, have imposed significant challenges on the economic returns of fresh fruit production in the present era. The fruit harvesting period is concentrated, demanding a substantial amount of labor. At the current stage, the harvesting of fresh fruits still heavily relies on manual labor, resulting not only in diminished efficiency but also considerable physical exertion. Turbulence in the global landscape and the rampant spread of infectious diseases have led to a sharp decline in the mobility of agricultural laborers, contributing to a labor shortage in orchard fruit production. Consequently, this scarcity has intensified the economic pressures on the orchard fruit industry's profitability.

Modern agricultural equipment and technology, pivotal components for augmenting agricultural productivity and emancipating rural labor, assume crucial significance in alleviating the scarcity of agricultural labor and enhancing the efficiency of fresh fruit harvesting. Our team focuses on apple harvesting robots as the research subject, centering our study around the optimization of apple picking poses. This research encompasses equipment development and experimental verification involving interactions between end effectors and fruits, fruit recognition and localization, and mechanical arm control. Specifically, our team investigates:

1) The impact of different picking methods on fruit separation. We establish a finite element model of the branch-stem-fruit system to provide a theoretical foundation for end effector gripping, damage assessment, and picking action optimization.

2) The gripping capability of the flexible three-finger FRE structure using the finite element method. We evaluate the potential for fruit damage. Through variance analysis and response surface analysis, we explore the significant influence of soft finger material hardness, fruit center of mass to palm distance, and fruit size on gripping force, aiming to optimize the end effector structure.

3) Lightweight algorithms for recognizing target fruits in complex environments influenced by factors such as fruit obstruction, lighting, and oscillation.

4) The application of "human-like" picking actions on harvesting robots based on human hand motion capture data.

Currently, two apple harvesting robots have been designed. One utilizes a four-degree-of-freedom robotic arm with a suction-type end effector. The robot detects, positions, grips, detaches, and places apples. A manipulator controller was designed to ensure rapid control execution. Picking experiments were conducted in a spindle apple orchard. The picking success rate of the rotation-pull pattern was 47.37% in the field orchard and 78% in the simulated orchard environment, with a picking cycle time of approximately 4 seconds. The stem damage rate in the field orchard was 11.11%. The other robot integrates a six-degree-of-freedom robotic arm with a flexible three-finger FRE structure. Orchard trial results demonstrate an 80.17% success rate in harvest using humanoid-inspired picking methods. These studies lay a robust foundation for achieving mechanized fresh fruit harvesting in the future.